Velocity obstacle guided motion planning method in dynamic environments

Abstract

The ability of a robot to navigate through a dynamic environment, avoid obstacles, and reach its destination is crucial for safety and a major technological challenge in autonomous navigation, especially in crowded environments involving hospitals, hotels, and restaurants. Most prominent methods use distance metrics as a safety constraints in their planning algorithms, which leads to overly cautious navigation. Thus, a velocity obstacle guided motion planning method for mobile robots in moving obstacle environments is proposed. The approach automatically negotiates dynamic obstacles combining velocity obstacle (VO) and trajectory generating. First, a dynamic perception algorithm is developed to track and predict obstacles using only point cloud input. Subsequently, to obtain an initial trajectory that satisfies kinodynamic and the VO-based safety metric (VOSM), VO is applied to search the kinodynamic path to verify the safety of the extended motion primitives. Finally, the smoothness and safety of the robot trajectory are enhanced by nonlinear optimization, which incorporates the proposed safety constraints based on VO. Extensive simulations in challenging environments demonstrate a 40.0% success rate increase and a 44.8% trajectory smoothness improvement over obstacle distance-based methods. Practical testing prove our technology is reliable and can safely avoid dynamic obstacles.